Skills I've built for Claude

A nightly enforcer that makes sure none of my code is stranded on my laptop.

Scans my machine for git repos with no remote and directories with unpushed commits, auto-creates private GitHub repos for anything unsynced, pushes the pending work, gitignores secrets, and emails me what it did. Runs every night at 11:15pm.

Broken deploys fix themselves before I've read the failure email.

Scans Gmail for Vercel deployment-failure notifications, pulls the repo, reads the build and runtime logs, diagnoses the error, patches it, and redeploys. Runs every three hours during the day.

Bug reports on my live interior-design site turn into shipped fixes.

A webhook fires the moment I submit the in-site bug form. The skill reproduces the bug, writes the patch, and deploys it — the report and the fix are the same loop.

Supply-chain security for the AI skills I install.

Vets, installs, and monitors third-party Claude skills through a five-stage pipeline — provenance manifest, hash baselines, sandbox checks, and a weekly drift scan that flags anything that changed underneath me.

The engine behind every professional message I send.

Drafts cold emails, LinkedIn DMs, intro requests, and referral blurbs in my voice, and tailors my resume to a specific role — one system for networking, outreach, and applications.

Finds the roles and the people before they hit the job boards.

Runs natural-language job search, Google X-Ray across Greenhouse / Lever / Ashby, scouts companies showing growth signals, and surfaces the likely hiring manager for a role.

My application pipeline, always current.

Tracks every company, resume status, outreach sent, and application submitted, so I always know what's outstanding — and so parallel Claude sessions don't collide on the same data.

Tailors and publishes my resume for a specific role.

Reads the job description, rewrites bullets to match, edits the Google Doc, and publishes a per-role copy to schlacter.me — the full loop end to end.

Teaches the resume skill so I never give the same note twice.

Routes feedback from a hands-on editing session back into the canonical resume files, editing them in place instead of spawning duplicate docs.

My interview prep and mock-interview coach.

Researches the company, maps my stories to the role, runs behavioral and product-sense mocks, and gives feedback on my answers.

Turns a job application into a real prototype.

Scopes and builds a targeted, vibe-coded project that demonstrates value to a specific team — the same approach behind several case studies on this site.

A stylist that knows my closet and a thousand outfits.

Searches a database of 1,000+ analyzed influencer looks from Instagram, LTK, and ShopMy, cross-referenced with my own wardrobe inventory, to answer “what do I wear with this?”

A design advisor that already knows my house.

Holds the full context for our Oakland home — style, color palette, room priorities, team contacts, and specs — so I can ask a design question without re-explaining the whole project.

Keeps the house's inspiration gallery in sync.

An image pipeline that moves design inspiration into the right categories in my interior-design portfolio, so the gallery stays organized without manual filing.

My writing voice, encoded as a system any task can layer on.

A five-level register model and a fourteen-point checklist that any writing task — emails, docs, applications, this page — runs through, so the output reads as mine instead of generic.

An AI-slop detector that catches the tells before I ship them.

Scans a draft against a twelve-category checklist for generic, jargon-heavy, hedge-everything AI writing, then rewrites the flagged lines in my voice. A required step before anything goes out.

A thinking partner that pushes back.

A DM-style sparring partner — short bursts, strong opinions with tradeoffs, one question per turn, no premature conclusions. For thinking out loud on anything before I converge.

Raises the floor on code quality by never trusting a single first draft.

Generates five parallel implementations of the same task, self-scores each against a measurable rubric, and merges the winner. Exercises judgment on when a task actually warrants the full five-draft pass.

Contributing back to the protocol the rest of these skills run on.

Guides upstream contributions to the Model Context Protocol governance org — the spec, the official SDKs, and the docs — including the full SEP proposal lifecycle from triage to sponsor review.

Turns a working prototype into case-study assets with zero manual steps.

Drives macOS screencapture, ffmpeg, and headless Chrome to produce clean stills, screen recordings, and GIFs — the same pipeline that captured the demos across this site.

The skill that maintains this site.

Knows the structure, design tokens, and deploy flow of schlacter.me, so adding a project or shipping a change is one instruction instead of a hunt through the codebase. This page was built with it.

Reddit search and research, on tap.

Searches posts and subreddits and pulls full discussions — the data layer behind the product-research dashboards on this site.

Reads a YouTube video so I don't have to.

Pulls metadata and the full transcript from any video through a credential-free pipeline, so I can summarize it or take structured notes.

Direct control of my Google Docs.

A skill I built over the Google Docs API to read, create, edit, search, and format docs — including the careful find-and-replace flow my resume workflow depends on.

Fast, reliable Google Drive from a prompt.

Searches with Drive query language, reads and downloads files, and uploads new ones — mapped around the API's quirks so it just works.

My own usage data, aggregated and made public.

Parses my local Claude Code transcripts into a live stats page — hours, sessions, commits, top projects — that updates weekly. A PM claim about myself, with the number on the page.

A weekly self-review loop that rewrites its own rulebook.

Reads my session transcripts, writes per-session post-mortems, extracts the habits that keep recurring, and edits my global instructions so future sessions correct for them automatically.

Clean handoffs between sessions.

Stops work and writes a complete session summary so I can close a session and a fresh Claude picks up exactly where I left off — the reason my long projects survive context limits.